12(34)
their competitors. This means that a rational, profit-maximizing producer
should put forward in its tender a bid that is close to its actual valuation
of the project.
As a consequence, it will be assumed that producers reveal their true
valuations when submitting their tenders and that a classical supply curve
can thus be derived from the tenders. The number of tenders will also
affect the price level in a non-strategic setting, resulting in a lower
expected price as the number of tenders increases.
Given that the distribution functions are estimated based on tenders
actually presented, it is necessary to ask whether there is a risk of selection
bias, that is, whether the supply function derived from actual tenders
might be skewed in some direction. It is sometimes claimed that small
and medium-sized enterprises are treated unfairly by the procurement
framework, in which case the estimates would yield a distorted picture of
the underlying distribution.
The Swedish National Financial Management Authority has investigated
this problem by studying framework agreements among the approximately
50 000 suppliers to central government (ESV 2008). The result is that there
are no signs of discrimination against small and medium-sized enterprises.
The dominant groups of suppliers in framework agreements in 2007 were
enterprises with 10–19 or 20–49 employees. As a consequence, we have
no reason to believe that the tenders presented represent a biased sample
of the underlying distribution.
It should be added that the outcome of the analysis is not critically
dependent on small and medium-sized enterprises being correctly
represented in proportion to their actual prevalence among enterprises
in general. The procurement routine may work well even if they are not,
provided that the barrier to entry is not prohibitive.
In accordance with the above arguments, the procurement operations
forming the basis of this study were analyzed by using a non-strategic
description. Distribution functions were estimated from the tenders for each
procurement operation, as a basis for the counterfactual analysis of what
would happen in the absence of a full-scale open procurement. As a test
of the robustness with respect to the assumption of non-strategic behavior,
consider what would be the difference between the models if we were to
assume that the number of tenderers and their valuations were in fact
known before the tenders are presented. According to the previously cited
formula, the optimal price level in an auction is given by
b*(v) = v – ∫ F(x)
(n-1)
dx / F(v)
(n-1)
.
The number
v
is the actual valuation, so the difference between the non-
strategic and the strategic models is simply the second term on the right-
hand side (with the sign reversed in the procurement situation). In the
population under study, the average number of tenders was between 5 and
6. Given that this number was not known and that uncertainty would lead
risk-neutral tenderers to use a higher number if they behaved strategically,
it is appropriate to choose a higher number, say between 7 and 8, as the
benchmark for a typical procurement operation. If the distribution function
is approximated by a polynomial for
v
-values less than v
min
, it turns out
that the correction term amounts to 4 or 5 percent of v
min
. Further, if we
